Skip to content
/ ploi Public

Code for "Planning with Learned Object Importance in Large Problem Instances using Graph Neural Networks" (AAAI 2021)

License

Notifications You must be signed in to change notification settings

tomsilver/ploi

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PLOI

This repository houses code for the AAAI 2021 paper:

Planning with Learned Object Importance in Large Problem Instances using Graph Neural Networks

Tom Silver*, Rohan Chitnis*, Aidan Curtis, Joshua Tenenbaum, Tomas Lozano-Perez, Leslie Pack Kaelbling.

For any questions or issues with the code, please email [email protected] and [email protected].

Link to paper: https://arxiv.org/abs/2009.05613

Link to video: https://www.youtube.com/watch?v=FWsVJc2fvCE

Instructions for running (tested on Mac and Linux):

  • Use Python 3.6 or higher.
  • Download Python dependencies: pip install -r requirements.txt.
  • Make sure the directory containing this code is in your PYTHONPATH.
  • Download and build the plan validation tool available at https://github.com/KCL-Planning/VAL, then make a symlink called validate on your path that points to the build/Validate binary, e.g. ln -s <path to VAL>/build/Validate /usr/local/bin/validate. If done successfully, running validate on your command line should give an output that starts with the line: "VAL: The PDDL+ plan validation tool". If you have trouble with the symlink, you can just directly change VALIDATE_CMD in planning/validate.py to point to the build/Validate binary. Note: we have found more success directly downloading binaries from https://dev.azure.com/schlumberger/ai-planning-validation/_build?view=runs (click the latest green run, click one of the Jobs near the bottom, then click "artifacts produced" to get to the downloadable binaries), rather than building from the Github source.

Now, ./run.sh should work. Different domains and methods can be run by modifying the variables at the top of run.sh.

About

Code for "Planning with Learned Object Importance in Large Problem Instances using Graph Neural Networks" (AAAI 2021)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published